Winding a transformer starts with your core material.
Unlike many RF transformers, it's not uncommon for power transformers to be composed of laminated steel plates. You'll first have to identify whether you have a ferrite material, or laminated steel plates. A given material will have a permeability which expresses a quantity of reluctance to a magnetic field, which tells you that for a specific number of turns, you'll get a particular amount of inductance.
Suppose you were using #77 core material for an E-core -500 set (
http://www.amidoncorp.com/ea-77-500/) - The spec sheet for this core (
http://www.amidoncorp.com/product_images/specifications/2-40.pdf) outlines 4470mH for 1000 turns (Al value).
Set that information aside for a moment.
The second thing you need to comprehend is that a transformer is a
power device, not a voltage device. Voltage is a component of power. So is current. Recall that you have a resistance when you have V/I? Well transformers convert impedances in their power transformation operation. You specified that you want 24V with a 3A load; or with an 8 ohm load on the output, you want 24V supply.
Power out = 24 * 3 = 72 Watts
Your input voltage is 180V in amplitude for 120Vrms line supply. Figure your transformer is pretty efficient at 98% Work your current backwards:
Pin = 72/0.98 ~= 73.5W
73.5 = 180 * InputCurrent
InputCurrent = 0.41A
Resolve your effective input Z: 441Ohms.
Your turns ratio? sqrt(441/50) [pri/sec] = 7.42:1
Now, "how many turns?" - this is dictated by your primary side magnetics; the "L" value will be the magnetizing inductance of your transformer, and you will typically want the impedance to be at least 5x your specified value, preferably 10x. (As you get closer to 1x, your transformer starts becoming more of a parasite and your coupling goes down)
Zeffective_primary: 441 x 10 = 4410 Ohms.
Now you have a primary impedance target: 4.4kOhm, so solve the inductance:
4410 / (2 * 3.1415* 60Hz) = 11.7 H
If you go for the 5x factor: 5.8 H
So, how many turns? It goes back to the Al number for your particular core. If you were married to the -500 core, your Al = 4470mH/1000T
How many turns is 5.8H?
Num_turns = 1000 * sqrt( 5800mH/4470mH ) = 1140 Turns.
How many turns is 11.7H?
Num_turns = 1000 * sqrt( 11700mH/4470mH) = 1618 Turns
From here, you can resolve your secondary turns:
1140 T / 7.42 = 154 turns secondary
1618 T / 7.42 = 218 turns secondary
That's a lot of turns isn't it?
Well, the thing to understand that just as Capacitance is a function of Electric Field for volume of space (thickness and area), Inductance is a function of magnetic field for a given volume. Those really,
really big transformers? They have an even larger Al value (often being a few 10s of H per 1kTurns).
So now you've done all the math work, and you have your basic numbers. Now comes the materials challenge: For the given number of turns you have calculated, for the given frequency of constraint, and the applied voltage available to your source, you have to calculate flux density for your core material. This is incidentally why laminated steel is popular for power transformers, as it has a very high flux capability before becoming saturated. Good manufacturers will provide the general equations for the calculation of the Bl quantity and give you a number to stay under. But since this is highly dependent on your source, you'll have to either get answers from your vendor, or calculate and measure where the core saturation hits.
Spec'ing a transformer design is actually an iterative process, like many things in Engineering.
Once you've completed your transformer, you should validate its basic operation with a function gen. Set it to 60 Hz, with a sine-wave output at max voltage. With the 50ohm source, your secondary impedance should be 0.9 Ohm. Fudge and use a 1Ohm resistor. If things are correct, for 10V input amplitude, you should resolve 1.347V output amplitude.
Once you have a basic verification of operation, you would be ready for testing it on a "live" system, however you would want to use an isolation transformer, with an inline fuse, to cover your butt in case you did something wrong. Put an 8ohm load on your secondary output (rated to 100W). Your initial test would be best performed with a remote operation so that if anything explodes, you're not near it. If it passes these tests, you'll next want to have the system running, and take thermal measurements of the device to ensure it does not get exceedingly hot. If it's getting hot, something is wrong and it isn't a good design (likely the flux density is much higher than calculated and you'll have to turn another prototype to get the losses down).
Safety is key, especially with power systems. Don't blame me if you screw up, but there is the information needed.
Cheers.